Abstract A decade and a half after the release of the human genome sequence, how well do we understand the molecular mechanisms underlying genotype-phenotype relationships? Importantly, as we generate ever- increasing numbers of genome sequences from diseased but also healthy individuals, how well can we predict the phenotype of an individual from information available about their genotype? The more we learn about the human genome, the further we seem to deviate from the simple model: ?mutation in gene X leads to perturbation of gene product X, which leads to disease A?. Among the most prevailing and mysterious deviations from this simple model are: (i) incomplete penetrance, whereby only a subset of individuals carrying a mutation are affected by the disease, and (ii) variable expressivity, whereby not all individuals affected by a given mutation are affected equally. These two interconnected phenomena have recently attracted attention because sequencing of exomes and genomes of healthy individuals shows an unanticipated burden of damaging mutations, with an average of ~300 damaging variants and >50 variants causing Mendelian disorders. This burden of damaging variants suggests a heretofore-unrecognized level of genetic resilience. It is becoming increasingly clear that gene products function in the context of complex interactome networks that need to be considered to fully illuminate genotype-phenotype relationships. Proteome-scale systematic interactome maps, which model these networks of molecular components and interactions between them as ?nodes? and ?edges?, respectively, are rapidly becoming available to initiate such approaches. We have started to dissect how disease-associated mutations impair interactions in the context of interactome networks. We have found that while common variants from healthy individuals rarely affect protein-protein interactions or DNA-protein interactions, a majority of disease-associated alleles perturb interactions, with about half corresponding to what we refer to as ?edge-specific? or ``edgetic'' alleles, i.e. alleles affecting a single or a subset of interactions while leaving other interactions unperturbed. This grant application is focused on understanding incomplete penetrance based on gene-gene interactions. Our central hypothesis is that incomplete penetrance can very often be best explained by edgetic alleles that are genetically suppressed by compensatory alleles in interacting partners. Due to a huge limitation in statistical power to test this ?edgetic suppression? hypothesis in human, we will first concentrate on S. cerevisiae as a model organism to develop the necessary concepts, methods and tools. We will leverage the recent release of ~1,000 yeast genome sequences to identify and characterize large numbers of pairs of natural variants that are damaging individually, but cooperatively functional. The strategies developed and general mechanisms discovered will be directly applicable to solving incomplete penetrance in humans.